"What amp input sees is 1V /314 roughly". Yes this is about 3mV input, more in the MM range. Keeping the input to 0.3mV will correlate better with what can be expected from a MC. In other words as far as what noise and distortion can be expected.
It is important to create simulations of real world conditions to better correlate the results with reality. As a case in point I attempted to use the GyroHead with MM cartridges. It technically "worked" but expectedly not so from a psycho-acoustic reality, seemingly generating distortions of the unwanted kind.
By the way, I am not keen on engaging overwhelming in simulations without engaging at least periodically with real world implementations. You can get way offtrack even though I love simulations.
It is important to create simulations of real world conditions to better correlate the results with reality. As a case in point I attempted to use the GyroHead with MM cartridges. It technically "worked" but expectedly not so from a psycho-acoustic reality, seemingly generating distortions of the unwanted kind.
By the way, I am not keen on engaging overwhelming in simulations without engaging at least periodically with real world implementations. You can get way offtrack even though I love simulations.
Last edited:
Ok, ok, I will reduce input from 1V to 100mV next time, just to be exactly at range, I doubt there will be much difference... Lets see
Hi,
in #150 I meant You might check what happens to the input potentials and bias currents when one supply line goes gonski.
It might happen that at least one of the inputs gets so much off of the nominal value that lethal currents may run through an attached pickup.
A blocking large-valued cap may be neccessar to protect the pickup.
jauu
Calvin
in #150 I meant You might check what happens to the input potentials and bias currents when one supply line goes gonski.
It might happen that at least one of the inputs gets so much off of the nominal value that lethal currents may run through an attached pickup.
A blocking large-valued cap may be neccessar to protect the pickup.
jauu
Calvin
Jauu,
It's in #158, for the moment good enough.
If real cart will be connected to this, I will return to subject to ensure its 100% safe
It's in #158, for the moment good enough.
If real cart will be connected to this, I will return to subject to ensure its 100% safe
Hi, I forgot to do this in post #158, so now update:
I got 108mV on Rg resistor (aka cart) regardless which side I grounded, also with +- 8V PS.
Since My Rg is 33 Ohm, and yours is 15, I think results correspond nicely from Canada to Croatia 🙂
For today's morning coffee I made 2 more tests just to clean up common setup understanding. Both still with preamp in current mode as #158
First, what I should have done earlier; comparing no signal inputs through preamp with soundcard shorted from out to in:
Only soundcard:
Through preamp:
You can see that pure closed loop without preamp is better than when going through preamp, not major difference but it will be difficult measuring super low noise like this.....
Second thing is clean 50Hz signal (plus harmonics; 100, 150, 200Hz and on) when going through preamp. Even I took big care with cables and sealed enclosure, as well as battery supply, still 50Hz is picked up. I wonder what will happen with it later when fully differential in and out will be used.
As for RME and its SMPS , hat down to engineers building it, no trace of 50 Hz!
First, what I should have done earlier; comparing no signal inputs through preamp with soundcard shorted from out to in:
Only soundcard:
Through preamp:
You can see that pure closed loop without preamp is better than when going through preamp, not major difference but it will be difficult measuring super low noise like this.....
Second thing is clean 50Hz signal (plus harmonics; 100, 150, 200Hz and on) when going through preamp. Even I took big care with cables and sealed enclosure, as well as battery supply, still 50Hz is picked up. I wonder what will happen with it later when fully differential in and out will be used.
As for RME and its SMPS , hat down to engineers building it, no trace of 50 Hz!
Second thing was reducing input signal to roughly 0.3 mV,
First I repeated as before, with approx 3mV
Comparing to exactly the same test in #158, today LF noise is much cleaner and 50 Hz artifacts become clearly visible.
Same thing with 10 times lower signal, approx 0.3 mV:
Looks as expected, all parameters proportionally worse, but still super. 50 hz artifacts clearly visible and higher than THD, as mentioned above whole set up is well shielded..... and it is still there... So much about why high CMMR...
First I repeated as before, with approx 3mV
Comparing to exactly the same test in #158, today LF noise is much cleaner and 50 Hz artifacts become clearly visible.
Same thing with 10 times lower signal, approx 0.3 mV:
Looks as expected, all parameters proportionally worse, but still super. 50 hz artifacts clearly visible and higher than THD, as mentioned above whole set up is well shielded..... and it is still there... So much about why high CMMR...
I haven't done it yet, though it seems the fault currents wouldn't exceed values obtained earlier from the topology shown, though I might check this later.
Yes, but if a network needed a capacitor I wouldn't necessarily be using it. The low impedances of a moving coil requires "high" value capacitance to get the low frequency bandwidth. That's the main reason I designed the GyroHead as there are no blocking or coupling capacitors. IMO capacitors of the electrolytic type can take days to sonically neutralize, making them unsuitable in a battery driven device. However under circumstances the equipment can be left on permanently electrolytic capacitors can neutralize, sometimes sonically preferable to polypropylene's.
For a dose of reality I connected the Denon 103R in the current mode across the emitters of the input stage of the SSM2019. Using a test record containing a 1KHz 5cm/sec sinusoid the output of the SSM2019 measured 324 mV RMS. This means that the gain of the network is about 62dB for a cartridge having an output of 250uV unloaded, or over x1000. In playing some normal records the output levels peaked at about +/- 400mV, suggesting that amplified higher frequencies created by the inverse RIAA weren't that high and manageable.
The GyroHead has only been posted in this thread. Besides for power it uses the battery as the feedback loop balancing element in a kind of folded long tailed pair configuration (thats why it is so named). This is why it requires two batteries, one for each channel. In this way the frequency response can extend differentially from DC into the megahertz region. The caveat is that it requires signals to be input floating differentially.
Nice, nice. I will do the same as soon as my TT is back in action..... could be early fall do, parts take long to come...
Aaaa, understand now, the name lead me to believe that there is more behind it. Thx for explaining.
Simple head amp looks very fine, even you lost me with battery as feedback loop, I just don't see it there (the feedback)... I see 2 x 1k resistors creating AC zero from outputs that feeds bias for transistors through 10k, and rest is LTP ? Might-be i don't look with right eyes?
PS, Wiring cartridge floating should not be caveat, One playing with such stuff should know how his arm is wired. I see also commercial differential preamps that require their not diy customers to take TT to shop for check and rewire if needed...
Aaaaah, so I made one really bad mistake, all the measurements I took and show you here I did with 0Hz to 100kHz bandwidth, aggggh.
I was too eager to move on and solder something.... typical. Did not check all the settings after last play.
So sorry for that, I just noticed this mistake, also nobody that was looking at this spotted it (or at least did not drew my attention on it) .
In any case the measurements are not useless since diagrams and THD values are correct, but SNR is way out since it was integrating noise up to 100kHz
So back to basic; here is an picture of my soundcard measuring itself in close loop, first with 20Hz - 20kHz bandwidth (SNR 99.3db) , second 0 - 100KhZ (SNR69.8 db).
Only noise figures are much different, the rest isn't affected:
I was too eager to move on and solder something.... typical. Did not check all the settings after last play.
So sorry for that, I just noticed this mistake, also nobody that was looking at this spotted it (or at least did not drew my attention on it) .
In any case the measurements are not useless since diagrams and THD values are correct, but SNR is way out since it was integrating noise up to 100kHz
So back to basic; here is an picture of my soundcard measuring itself in close loop, first with 20Hz - 20kHz bandwidth (SNR 99.3db) , second 0 - 100KhZ (SNR69.8 db).
Only noise figures are much different, the rest isn't affected:
Hi,
I was talking about at least three cases I know from personally, where some quite expensive MC-carts where fried due to the failing of one of the supply lines.
So far I never experienced such a loss myself ... and I build Phono stages with INA inputs over more than 30 years by now ... and it is one's free personal decision to decide about how much risk to take ... but at least one should/could make a informed decision.
jauu
Calvin
I was talking about at least three cases I know from personally, where some quite expensive MC-carts where fried due to the failing of one of the supply lines.
So far I never experienced such a loss myself ... and I build Phono stages with INA inputs over more than 30 years by now ... and it is one's free personal decision to decide about how much risk to take ... but at least one should/could make a informed decision.
jauu
Calvin
Sure... I too have built phono stages for perhaps 50 years now, one currently being used by my pseudo son-in-law. I suspect I have burnt out a coil or two in my lifetime, though no moving coils that I can remember. More damage wearing them out, bending the cantilevers to the side or bending them in half. I have a couple of original Denon 103's, one with, and one without a stylus. The one without the stylus is a potential candidate to see if the coil will burn out with fault currents.
In testing with power supply collapse (or turn-on), the results were mostly as expected, with coil fault currents being found considerably less than grounding one side of the coil. Currents were simulated by a 15 Ohm resistor connected between the input emitters in current mode (bases grounded).
For supplies set to +/-15 volts current through the coil measured about 0.1mV or 0.1mV/15 Ohm = 6.7uA
With +15 V open circuit (loss of power supply rail) the coil measured 16.6mV = 1.1mA
With +15 V grounded the coil measured 0.3mV = 20uA
With -15 V open circuit the coil measured 0.6mV = 40uA
With -15 V grounded the coil measured 1.8mV = 120uA
For +/-15 volt supplies
With pin 1 grounded the coil measured 53.6mV = 3.6mA
With pin 8 grounded the coil measured 44.0mV = 2.9mA
Setting supplies to +15/-5 volts current through the coil again measured 0.1mV =6.7uA
With +15 V open circuit the coil measured 8.5mV = 570uA
With +15 V grounded the coil measured 0.0mV = 0uA
With -5 V open circuit same as above = 40uA
With -5 V grounded same as above = 120uA
For +15/-5 volt supplies
With pin 1 grounded the coil measured 23.6mV = 1.6mA
With pin 8 grounded the coil measured 13.8mV = 0.92mA
Under normal circumstances the positive supplies are unlikely to be open circuit. (hence not going to the negative supply rail). This means the differential coil currents are unlikely to be much greater than about 120uA under any circumstances of supply voltage changes. In contrast connecting one of the ends of coil to ground from the normal -0.6 volt emitter voltage causes coil currents to reach up to 3.6mA. This can be reduced to less than half (1.6mA) by using a -5V supply rail in conjunction with a +15 V or +5 V positive supply.
Coil fusing relates to reaching a fusing temperature that is dictated by the inability to dissipate power in degrees C/watt. Meaning that fusing relates in part to I^2/R, that the power requiring dissipation rises as the square of the difference in fault current. Although the current reduces from 3.6mA to 1.6mA the power causing temperature rise is about 1/5 for a 5 Volt supply. This also means that higher resistance coils will heat faster, as fault currents are largely independent for low coil resistances.
It should be realized that the power requiring dissipating for a fault current of 1.6mA is 1.6mA^2 x 15 Ohms = 38uW. Don't know if this is too much, though it doesn't seem likely it would fuse.
In testing with power supply collapse (or turn-on), the results were mostly as expected, with coil fault currents being found considerably less than grounding one side of the coil. Currents were simulated by a 15 Ohm resistor connected between the input emitters in current mode (bases grounded).
For supplies set to +/-15 volts current through the coil measured about 0.1mV or 0.1mV/15 Ohm = 6.7uA
With +15 V open circuit (loss of power supply rail) the coil measured 16.6mV = 1.1mA
With +15 V grounded the coil measured 0.3mV = 20uA
With -15 V open circuit the coil measured 0.6mV = 40uA
With -15 V grounded the coil measured 1.8mV = 120uA
For +/-15 volt supplies
With pin 1 grounded the coil measured 53.6mV = 3.6mA
With pin 8 grounded the coil measured 44.0mV = 2.9mA
Setting supplies to +15/-5 volts current through the coil again measured 0.1mV =6.7uA
With +15 V open circuit the coil measured 8.5mV = 570uA
With +15 V grounded the coil measured 0.0mV = 0uA
With -5 V open circuit same as above = 40uA
With -5 V grounded same as above = 120uA
For +15/-5 volt supplies
With pin 1 grounded the coil measured 23.6mV = 1.6mA
With pin 8 grounded the coil measured 13.8mV = 0.92mA
Under normal circumstances the positive supplies are unlikely to be open circuit. (hence not going to the negative supply rail). This means the differential coil currents are unlikely to be much greater than about 120uA under any circumstances of supply voltage changes. In contrast connecting one of the ends of coil to ground from the normal -0.6 volt emitter voltage causes coil currents to reach up to 3.6mA. This can be reduced to less than half (1.6mA) by using a -5V supply rail in conjunction with a +15 V or +5 V positive supply.
Coil fusing relates to reaching a fusing temperature that is dictated by the inability to dissipate power in degrees C/watt. Meaning that fusing relates in part to I^2/R, that the power requiring dissipation rises as the square of the difference in fault current. Although the current reduces from 3.6mA to 1.6mA the power causing temperature rise is about 1/5 for a 5 Volt supply. This also means that higher resistance coils will heat faster, as fault currents are largely independent for low coil resistances.
It should be realized that the power requiring dissipating for a fault current of 1.6mA is 1.6mA^2 x 15 Ohms = 38uW. Don't know if this is too much, though it doesn't seem likely it would fuse.
Last edited:
Dears,
So i start again (this time with correct bandwidth 😎 ), SSM2017 in current mode (bus summing) , Rg 33 Ohm, Rin 2 x 47k.
Input from soundcard is 1V reduced by 47k and 33 for approx 2850 times to bring it on MC cart level
Output from Amp : 314 mV, gain approx x100 or 40db
SNR is now -77,3 db, THD 0.0012%. Now it looks right.
In my opinion above will be masked by even the best cartridge - TT - LP combo and make it irrelevant... Such circuit can go in as it is, SSM2017 is great...
But I'll continue.
PS, for some reason also 50 Hz noise vanished today????
So i start again (this time with correct bandwidth 😎 ), SSM2017 in current mode (bus summing) , Rg 33 Ohm, Rin 2 x 47k.
Input from soundcard is 1V reduced by 47k and 33 for approx 2850 times to bring it on MC cart level
Output from Amp : 314 mV, gain approx x100 or 40db
SNR is now -77,3 db, THD 0.0012%. Now it looks right.
In my opinion above will be masked by even the best cartridge - TT - LP combo and make it irrelevant... Such circuit can go in as it is, SSM2017 is great...
But I'll continue.
PS, for some reason also 50 Hz noise vanished today????
It should be noted that the above spectrum is pre-RIAA, hence harmonics and noise are attenuated by the imposition of an RIAA by about 6dB/octave. In contrast hum and lower frequency difference frequencies caused by intermodulation distortion rises as frequency goes down. Hence the above requires some interpretation in relation to other networks with flat frequency response that are post RIAA.
I did some further tests on the SSM3019 to determine bandwidth and gain/bandwidth product using the above network. Signal was applied to the normal input, being attenuated by about 1000:1 with low Ohmic source impedance of 10 Ohm to the base. With a simulated coil resistance of 15 Ohm the input to output was 2.39V RMS out for 3.9V RMS in. The SSM2017 gain was therefore about 0.612x 1000 or about 56dB.
In absence of the 5n6 capacitor the 3dB bandwidth was 330KHz. With 5n6 included the bandwidth extended flat (non-peaking) to about 400KHz. The gain/bandwidth of the network can be calculated of equivalency at 330KHz x 612 or about 200MHz GBP and to 400KHz x 612 or 245MHz GBP with 5n6 capacitance. In contrast the NE5532 has a unity gain bandwidth of 10MHz. In this regard the SSM3019 seems formidably talented. The inclusion of the capacitance has application to filter out higher frequencies into the RFI region by shorting the emitters together to prevent rectification in becoming an AM detector. With the bases also shorted (to ground) little signal can pass into the secondary differential stage.
Another test conducted was to measure offset variances by the application and removal of higher frequency (400KHz) level signals with frequency, consisting of 3.9VRMS being turned on and off. The result was very good in that the DC offset didn't vary much from about 70mV, a value that can ultimately be balanced out in some way. Although expected to be reasonably good it is often the case that imbalanced amplification stages with square law transfer functions shift the DC point, whereupon intermodulation products are folding down. Because of the post RIAA network the low frequencies can be dramatically amplified, causing the cones of the woofers to extend more so than safely desired.
I did some further tests on the SSM3019 to determine bandwidth and gain/bandwidth product using the above network. Signal was applied to the normal input, being attenuated by about 1000:1 with low Ohmic source impedance of 10 Ohm to the base. With a simulated coil resistance of 15 Ohm the input to output was 2.39V RMS out for 3.9V RMS in. The SSM2017 gain was therefore about 0.612x 1000 or about 56dB.
In absence of the 5n6 capacitor the 3dB bandwidth was 330KHz. With 5n6 included the bandwidth extended flat (non-peaking) to about 400KHz. The gain/bandwidth of the network can be calculated of equivalency at 330KHz x 612 or about 200MHz GBP and to 400KHz x 612 or 245MHz GBP with 5n6 capacitance. In contrast the NE5532 has a unity gain bandwidth of 10MHz. In this regard the SSM3019 seems formidably talented. The inclusion of the capacitance has application to filter out higher frequencies into the RFI region by shorting the emitters together to prevent rectification in becoming an AM detector. With the bases also shorted (to ground) little signal can pass into the secondary differential stage.
Another test conducted was to measure offset variances by the application and removal of higher frequency (400KHz) level signals with frequency, consisting of 3.9VRMS being turned on and off. The result was very good in that the DC offset didn't vary much from about 70mV, a value that can ultimately be balanced out in some way. Although expected to be reasonably good it is often the case that imbalanced amplification stages with square law transfer functions shift the DC point, whereupon intermodulation products are folding down. Because of the post RIAA network the low frequencies can be dramatically amplified, causing the cones of the woofers to extend more so than safely desired.
Hi, I'm very happy you are also dissecting this chip, its worth it to learn more about it in my opinion. Looking forward for more!
All correct, this is just first stage before any EQ, . Second stage after RIAA will also have challenging task, but we are not there yet.